Britta Allgöwer and Reto Schöning
By Swiss federal law (Nationalparkgesetz 1980) the Swiss National Park anticipates the complete protection of nature and its processes. Natural Forest fires belong to the ecosystem of forests. But the Swiss National Park (Engadin valley) is close to human settlements, it is small (170 sqm) and it is - also by law - liable for damages caused by it. In order to fulfil their assignment the park authorities need a fire management policy which allows the individual treatement of each fire situation and which includes ecological as well as economical aspects in the decision.
As forest fires are spatial processes Geographical Information Systems (GIS) are applied as tools to achieve an operational forest fire management system. Since the Division of Spatial Data Handling (University of Zurich) runs the GIS of the Swiss National Park, the people in charge of it work on three topics: (1) implementation of fire spread modelling in Geographical Information Systems, (2) development of fuel models for Switzerland and introduction to the spread modelling and (3) concepts for forest fire management strategies with special emphasis on protected areas.The overall goal is to provide an interactive Decision Support System which enables the training in a "Fire Simulator" of the people in charge (park authorities, local fire fighters) and the prediction of the damage potential.
The basis for the fire behaviour modelling is the Rothermel model for the behaviour of surface fires (Rothermel 1972). It calculates for any given point local intensity and spread parameters for the head of a surface fire. Inputs for the model are a two-dimensional wind field, terrain parameters, fuel moisture and a detailed description of the fuel bed. Based on the local behaviour output by the Rothermel model and on a model for the local shape of fire spread (Anderson 1983), the spread from a set of source locations can be simulated. The influence of barriers (streets, rivers, fuel breaks, etc.) is addressed with a probabilistic model based on the width of the barrier and the flame length. The spread simulation also allows the calculation of the flame length on the entire fire perimeter, which in turn is an important index for the success of various types of fire suppression activities (Rothermel 1983). Once all the required data is available for the Swiss National Park, the model can be used to evaluate different climatic and management scenarios. The fire spread model is implemented in SPARKS, a prototype fire behaviour modelling application. It is fully integrated in a commercial Geographical Information System (ARC/INFO), built on its raster modelling and applications development functionalities. This allows not only for synoptic analysis over very large areas, but it enlarges greatly the capabilities of the modelling package through the availability of the full range of the GIS' database and spatial analysis functions
Through the integration of fire behaviour models with GIS models, new insights in the fire danger situation in a management area can be gained. One example is the damage potential that arises from fires starting at a certain point in the landscape. This potential clearly depends on the proximity of the point to sensitive objects and areas like buildings, railway lines, fire-sensitive ecosystems, etc. Proximity is a concept which is used in a great many GIS-related models. However, in the mentioned example proximity can not be modelled as straightline distance, but it must take into account the behaviour of the fire spreading over the landscape. In this approach, the spread simulation is used to calculate the time it takes a fire starting from any point in the landscape to reach an object, under given environmental conditions. This is accomplished by inverting the spread simulation, working from a reached object backwards to all possible sources. The delay times from any point to all objects can then be input to a potential model, used in the GIS realm for assessing accessibility. The model weighs the influence of any reached object on the point's damage potential based on the delay time and the damage susceptibility of the object. The index for the fire damage potential arising from the point is then obtained by simply adding the weighted influences of all objects. This index could be further combined with fire occurrence estimations, probability for early detection, accessibility etc. to give a more complete image of the fire danger situation.
Many of the input parameters for the fire behaviour model must be modelled or gathered in extensive field surveys. In order to allocate resources required for the collection of these inputs and to assess the uncertainty introduced in the model results due to uncertain inputs, sensitivity and error analysis with Monte-Carlo Simulation can be performed. This allows the examination - in tabular form or graphically - of the relative importance of each input parameter for a selected output. Also, the uncertainty in the calculated fire behaviour can be calculated for interactively selected points, based on estimated uncertainties of the input parameters.
Anderson, H. E. (1983) Predicting Wind-Driven Wild Land Fire Size and Shape. Research Paper INT-305. Ogden, UT: US Department of Agriculture, Forest Service, Intermountain Forest and Range Experiment Station, pp. 1-26
Nationalparkgesetz (1980) Bundesgesetz über den Schweizerischen Nationalpark im Kanton Graubünden (Nationalparkgesetz) vom 19. Dezember, pp. 1-3
Rothermel, R. C. (1972) A mathematical model for predicting fire spread in wildland fuels. Research Paper INT-115. Ogden, UT: US Department of Agriculture, Forest Service, Intermountain Forest and Range Experiment Station, pp. 1-40
Rothermel, R. C. (1983) How to predict the spread and intensity of forest and range fires. General Technical Report INT-143. Ogden, UT: US Department of Agriculture, Forest Service, Intermountain Forest and Range Experiment Station, pp. 1-53